No Need to Dither?

… and this is something that we should have already been thinking about as errors, so I suppose it just goes to show that the availability of high-quality tools is no indication that the result will be high-quality output. EBU R 128 solves this as well, however in practice the goalposts are still moving a little bit, and in the end, until something exists whereby the non-observance results in lost revenue, it’s all just academic.

Personally I’d like to see an equivalent standard applied to CD and DVD such that duplication houses would not accept masters that are so clearly artificially “pumped”.

Anyway, this is possibly going a little bit off-topic. Personally, I would always start with a 32-bit float master and derive all other formats directly from that. Going to MP3, I have never found any need to dither because, let’s face it, the conversion is going to do far more damage than a lack of dither.

That’s the way I would do it too. For whatever reason, Apple won’t accept 32 bit float for MFiT. Makes no sense to me because they say their first step is 32 bit float (the caf file). But if the record label needs to make use of the 24 bit file for other things, I guess that would be their reason.

Yes, many MP3 albums are made from 16-bit 44.1 masters but, as Mr Soundman mentions above, that does not make it right. So just to clarify, my comments were recommendations about what it is best to do to get the best results. The concensus is that you get the best MP3 results going directly from a non-dithered high resolution 32-bit or 24-bit master file. And if your master file has to be dithered it’s best if it is flat TPDF dither (no noise shaping). And, of course, you need to avoid hyper compressed levels / inter sample peak distortion on the master file - all the clipping on those MP3 files is likely due to the hyper compressed levels of the CD masters rather than due to the dithering.